Assisted Reproductive Technology and Birth Defects Among Liveborn Infants in Florida, Massachusetts, and Michigan, 2000-2010.

We used data from the States Monitoring ART (SMART) Collaborative, a consortium of participating states and the Centers for Disease Control and Prevention Division of Reproductive Health that promotes state-based surveillance and research on the maternal and child health outcomes of ART. 19 We used linked ART surveillance, birth certificates, and birth defects registry data for Florida, Massachusetts, and Michigan from 2000 to 2010. These states linked their birth defects registry information with birth certificate data and provided deidentified linked data sets to the Centers for Disease Control and Prevention. Then, data from the Centers for Disease Control and Prevention’s National ART Surveillance System were linked with the state vital records information using a probabilistic method. Maternal and infant date of birth, plurality, maternal residence zip code, and gravidity (live birth order plus pregnancy losses) were primary linkage variables. 20 Ancillary information such as maternal race/ethnicity, infant sex, and infant birth weight were used to resolve duplicate links. The overall linkage rate was 90.5%. Is assisted reproductive technology associated with an increased risk of birth defects? In this cohort study of more than 4 million liveborn infants in 3 states, the prevalence of nonchromosomal birth defects among singleton infants was higher for those conceived using assisted reproductive technology compared with non–assisted reproductive technology singletons. This difference was significant after controlling for maternal characteristics. Infants conceived after assisted reproductive technology had a higher prevalence of certain birth defects; however, this association could be owing to underlying subfertility. We included all resident live births in Florida, Massachusetts, and Michigan between 2000 and 2010; ART births were those that were successfully linked with the National ART Surveillance System. We excluded births with missing information on plurality and ART births using methods other than transcervical IVF such as gamete intrafallopian transfer and zygote intrafallopian transfer (<1%for all states). The unit of analysis was a liveborn infant. The study was approved by the institutional review boards at the Centers for Disease Control and Prevention, Florida Department of Health, Massachusetts Department of Health, and Michigan Department of Health and Human Services. Informed consent was waived because the research involved no more than minimal risk, the rights and welfare of the subjects were not adversely affected, and the research could not practicably be carried out without such a waiver. The methods for ascertaining birth defect cases varied by state. The Florida Birth Defects Registry uses passive case-finding methods and ascertains birth defects diagnosed in liveborn infants before 1 year of age. Records are identified from hospital discharge abstracts, the state regional perinatal center database, and the state Children’s Medical Services records and are linked to birth certificates. The Massachusetts Birth Defects Monitoring Program uses active case finding whereby birth hospitals and pediatric care facilities submit discharge records with a birth defect diagnosis. Inclusion critera are being a liveborn infant or a fetal death of more than 20 weeks’ gestation or weighing more than 350 g, having a structural defect that meets the diagnostic criteria and was diagnosed before 1 year of age, and being born to mothers who were residents of the state at the time of delivery. The Michigan Birth Defects Registry uses passive case ascertainment based on reporting from hospitals, cytogenetic laboratories, and pediatric genetics clinics. Reporting is required for liveborn infants diagnosed with a reportable condition before the second birthday and for fetal deaths after 20 weeks’ gestation or more or after they reach 400 g or higher. Michigan data are augmented by information from Children’s Special Health Care Services enrollments; confirmed cases are identified as a result of newborn metabolic, hearing, or genetic screening and linked birth and death record information. Because of differences in case ascertainment across states, we evaluated a limited number of birth defects. We selected defects that were used in previously published national estimates 21 , 22 because they are usually diagnosed at or immediately after birth and are likely to be consistently collected and reported across different surveillance systems. The nonchromosomal birth defects assessed in this study were spina bifida with or without anencephaly, encephalocele, anophthalmia and microphthalmia, common truncus, transposition of great arteries, tetralogy of Fallot, atrioventricular septal defects, hypoplastic left heart syndrome, cleft palate without cleft lip, cleft lip with and without cleft palate, esophageal atresia/tracheoesophageal fistula, rectal and large intestinal atresia/stenosis, reduction defects of the upper limbs, reduction defects of the lower limbs, gastroschisis, omphalocele, and diaphragmatic hernia. The chromosomal defects were trisomy 13, trisomy 21 (Down syndrome), and trisomy 18. We examined the distribution of sociodemographic factors for all ART and non-ART liveborn infants. The variables were derived from the birth certificate and included maternal state of residence, age, race/ethnicity, education, parity (derived from live birth order), tobacco use during pregnancy, diabetes (chronic or gestational), hypertension (chronic or pregnancy-induced), infant plurality, sex of the infant, birth weight, and gestational age. We also compared the prevalence of birth defects for all ART and non-ART births and stratified by singleton and multiple births. For all comparisons, chromosomal defects were stratified by maternal age younger than 35 years and 35 years and older because of known age effects. 23 The prevalence of specific birth defects was reported if the defect was diagnosed in 20 or more infants. Next, we restricted the study population to ART-conceived infants and evaluated the prevalence of 1 or more nonchromosomal birth defects for certain ART procedures including cycle type (fresh nondonor, fresh donor, frozen-thawed nondonor, or frozen-thawed donor embryos), number of embryos transferred, use of assisted hatching (the purposeful disruption of an embryo’s zona pellucida by laser, mechanical, or chemical means), and infertility diagnosis (tubal factor, ovulation disorder, diminished ovarian reserve, endometriosis, male factor, or unexplained infertility). Because information on use of ICSI and day of embryo transfer was only available for fresh embryo cycles, we further restricted the study population to infants resulting from fresh embryo transfer and examined the prevalence of birth defects for all aforementioned procedures and ICSI with male factor infertility, ICSI without male factor infertility, and day of embryo transfer (days 2–3 vs days 5–6). We conducted the analyses for all live births and singleton live births; small numbers precluded evaluation of multiple live births. We used 2-tailed Satterthwaite-adjusted χ 2 tests to assess differences in the distribution of sociodemographic characteristics for ART vs non-ART liveborn infants. We used multivariable predicted marginal proportions from logistic regression models to compute adjusted risk ratios for the association between use of ART and birth defects. The models were adjusted for maternal age, race/ethnicity, education, parity, smoking during pregnancy, diabetes, hypertension, state of residence, and year of birth. We also used predicted marginal proportions to calculate adjusted risk ratios for the association between use of specified ART procedures and nonchromosomal birth defects. Models for births resulting from fresh and frozen-thawed embryo transfers included cycle type, number of embryos transferred, use of assisted hatching, and infertility diagnosis. Models for births resulting from fresh embryo transfers included type of ART (conventional IVF vs ICSI with and without male factor infertility), cycle type, number of embryos transferred, day of transfer, use of assisted hatching, and infertility diagnosis. Because information on maternal body mass index was not available for Massachusetts during the study period and was only available from autumn 2007 onward for Michigan and March 2004 onward for Florida, we were unable to evaluate this factor as a potential confounder. We applied the Holm-Bonferroni method to the primary and subgroup analyses to account for multiple comparisons and report multiplicity-adjusted P values. P values less than .05 were considered significant. All analyses accounted for clustering of infants within a live birth delivery and clustering of sibling births. We used SAS version 9.3 (SAS Institute) and SUDAAN version 11.0 (RTI International) for analysis. Cohort sizes of less than 20 infants were suppressed and complementary suppression was applied.

Between 2000 and 2010, there were 4 618 076 liveborn infants in Florida, Massachusetts, and Michigan; of those infants, 64 861 (1.4%)were conceived using ART. Compared with non-ART infants, those conceived using ART had higher frequencies of multiple births, a birth weight of less than 2500 g and less than 1500 g, and gestational age of less than 37 weeks and less than 32 weeks ( Table 1 ). There was no difference in the distribution of infant sex. Approximately 45.8%(n = 29 736) of ART-conceived infants were born to mothers who were Massachusetts residents, compared with 18.7%(n = 853 165) of non-ART infants. Mothers of ART infants were more likely to be 30 years or older, non-Hispanic white, college graduates, and nulliparous than mothers of non-ART infants. Tobacco use during pregnancy was less common in mothers of ART vs non-ART infants, while diabetes and hypertension were more common in mothers of ART infants. Among all liveborn infants, the prevalence of 1 or more selected nonchromosomal defects was 59.97 per 10 000(n = 389) for ART infants compared with 48.40 per 10 000 (n = 22 036) for non-ART infants ( Table 2 ). After adjusting for maternal characteristics and year of birth, ART use was associated with an increased risk for nonchromosomal birth defects (adjusted risk ratio [aRR], 1.28; 95%CI, 1.15–1.42). Tracheoesophageal fistula/esophageal atresia (aRR, 1.93; 95% CI, 1.40–2.67), rectal and large intestinal atresia/stenosis (aRR, 2.03; 95%CI, 1.51–2.74), and reduction deformity of the lower limbs (aRR, 2.18; 95%CI, 1.39–3.43) were positively associated with ART use. Among singleton live born infants, the prevalence of 1 or more selected nonchromosomal defects was 64.88 per 10 000 (n = 218) for ART infants compared with 48.07 per 10 000 (n = 21 251) for non-ART infants, and this association with ART remained significant in the adjusted model (aRR, 1.39; 95%CI, 1.21–1.59) ( Table 3 ). Likewise, the prevalence of tracheoesophageal fistula/esophageal atresia (aRR, 1.90; 95% CI, 1.23–2.94) and rectal and large intestinal atresia/stenosis (aRR, 1.88; 95% CI, 1.26–282) was higher for ART births vs non-ART births. For women younger than 35 years, the prevalence of Down syndrome was higher for ART vs non-ART births (aRR, 1.63; 95% CI, 1.05–2.54), but the association was not significant after accounting for multiple comparisons ( P = .18). For women 35 years and older, the prevalence of chromosomal defects was lower for ART births than non-ART births (aRR, 0.66; 95%CI, 0.49–0.88). With the exception of increased risk for rectal and large intestinal atresia/stenosis (aRR, 2.39; 95% CI, 1.38–4.12), no significant associations with ART were observed for multiple births. When the study population was restricted to ART births conceived by fresh or frozen embryo transfer, no significant associations between ART procedures and risk of 1 or more nonchromosomal birth defects were detected ( Table 4 ). After further restriction to ART births conceived by fresh embryo transfer, diagnosis of ovulation disorder was marginally associated with nonchromosomal defects among all live births (aRR, 1.53; 95% CI, 1.13–2.06; P = .05) ( Table 5 ). The prevalence of nonchromosomal defects was also increased for all live births (aRR, 1.32; 95% CI, 1.02–1.71) and singleton live births where assisted hatching was used (aRR, 1.55; 95% CI, 1.10–2.19); however, the multiplicity-adjusted P values were >.05.

Using data from a national ART surveillance system linked with vital records and information from 3 state-based birth defects registries, we found that ART use was associated with an increased risk for certain birth defects. These findings confirm previous reports based on smaller populations 3 – 5 and provide additional information on variations in risk according to the type of ART procedure used. While we did not find considerable differences in risk by procedure type, the prevalence of nonchromosomal defects was higher for ART births where assisted hatching was used compared with ART births without the technique; however, the association was not statistically significant after accounting for multiple comparisons. Two studies that assessed the risk of birth defects following assisted hatching did not detect an effect 14 , 15 ; however, such an association is plausible because the procedure could damage the embryo and is often used for patients with a poor prognosis who may have other risks for adverse birth outcomes. 24 In addition, we observed a marginally increased prevalence of nonchromosomal defects among births to women with an ovulation disorder, a finding that may be associated with high rates of obesity, a known risk factor for birth defects, 25 among women with polycystic ovary syndrome. 26 While we adjusted for diabetes in our models, it is possible that undiagnosed diabetes among obese women with polycystic ovary syndrome may partially explain this association. Overall, the prevalence of the selected birth defects in our study population is consistent with national estimates. 22 However, because we used a limited group of conditions, our prevalence estimates of 1% are lower than national estimates for all types of birth defects (approximately 3%). 27 In accordance with other studies, 3 – 5 , 7 , 28 – 32 we found that the risk of birth defects following ART varied by type of defect; the largest relative risks were observed for gastrointestinal and limb reduction defects. We also found an association between ART use and transposition of great vessels, which has been previously reported. 33 The consistency of our results with those of other studies and meta-analyses that assessed a broader group of birth defects suggests that the excess risk observed in our study is robust. The apparent negative association between ART and Down syndrome in women 35 years or older is probably because of the use of preimplantation genetic screening among older women, primarily for aneuploidy. 34 Notably, in women younger than 35 years, we found increased prevalence of Down syndrome among ART singleton infants compared with non-ART singletons. The reason for this association is unknown but may be caused by different attitudes toward pregnancy termination in younger women undergoing ART compared with women of the same age with spontaneous conceptions. Overall, older women tend to be more likely than younger women to terminate a pregnancy following a Down syndrome diagnosis 35 ; thus, there may be additional factors among younger women using ART that influence their opinions on termination. It is also possible that younger women with an ART-conceived pregnancy were less willing to undergo chorionic villus sampling or amniocentesis because of heightened concerns about risks to the fetus, leading to differences in the rates of prenatal diagnosis and consequent terminations. Another potential explanation is that young women undergoing ART have more serious underlying health issues than older women and thus have poorer-quality embryos. We did not find a significantly higher prevalence of selected nonchromosomal birth defects in ART births where ICSI was used vs those where conventional IVF was used after adjusting for patient and treatment characteristics. While this finding corroborates the results of meta-analyses published in 2012 and 2013, 6 , 7 our study did not include genitourinary defects, specifically, hypospadias and cryptorchidism, which are most often implicated in studies of ICSI and birth defects and which may be related to severe male factor infertility. 36 , 37 Similarly, we did not find an association between birth defects and use of donor oocytes or embryo stage at transfer; however, association with the transfer of 2 or more fresh embryos approached significance, particularly for singleton live births, suggesting that singletons originating from pregnancies where multiple embryos implanted may have increased risks for birth defects. Notably, the presence of a vanishing twin has been identified as a risk factor for small for gestational age in singleton births after IVF. 38 The primary strength of our study is the use of a large population- and registry-based cohort with accurate information on ART procedures. To our knowledge, this is the largest US study of birth defects and ART to date. Furthermore, we limited our analysis to birth defects that are apparent at birth and thus likely to be reliably ascertained across different states. However, our findings have several limitations. First, we did not have information on the occurrence of birth defects among fetal deaths or pregnancy terminations. As such, our prevalence estimates almost certainly understate the true prevalence of the birth defects included in our study. This may also result in risk ratios that are biased toward or away from the null, 39 depending on the extent to which mothers of ART infants are more or less likely than mothers of non-ART infants to have a miscarriage or terminate a pregnancy affected by birth defects. It is also possible that differences in case ascertainment across states influenced the combined prevalence estimates. Assisted reproductive technology–conceived infants may be monitored more closely, thus resulting in increased detection of birth defects in those infants compared with the general population. Because of variations among states in the collection of information on maternal body mass index, we were unable to control for this potential confounder. Finally, some of the birth certificate variables used in the adjustment are under- reported and may be differentially reported among ART and non-ART groups.

We found that ART use conferred an increased risk for nonchromosomal birth defects, particularly those affecting the gastrointestinal and musculoskeletal systems; however, we were unable to evaluate the potential effect of underlying subfertility on this association. Among ART births, no single procedure was found to substantially increase risk, although use of assisted hatching and diagnosis of ovulation disorders were associated with marginal increases in the prevalence of birth defects. In total, these findings suggest that factors related to subfertility may explain the association between use of ART and birth defects, although additional studies on specific ART procedures are needed. As use of ART continues to increase, careful evaluation of the long-term outcomes of children conceived using these technologies becomes increasingly important. These findings provide additional information on risks of ART that can be used when counselling patients.